Discovery of U-Carbon: Metallic and Magnetic Hong Fang1, Michael Masaki2, Anand B. Puthirath3, Jaime M. Moya4, Guanhui Gao3, Emilia Morosan4, Pulickel M. Ajayan3, Joel Therrien2*, Puru Jena1* 1 Physics Department, Virginia Commonwealth University, Richmond, VA 23284-2000, USA. 2 Department of Electrical and Computer Engineering, University of Massachusetts, Lowell, MA, USA. 3 Department of Materials Science and NanoEngineering, Rice University, Houston, TX, 77005, USA. 4Department of Physics and Astronomy, Rice University, Houston, TX 77005, USA. *Correspondence to: [email protected] (J.T.); [email protected] (P.J.) Abstract: We report the discovery of a pristine crystalline 3D carbon that is magnetic, electrically conductive and stable under ambient conditions. This carbon material, which has remained elusive for decades, is synthesized by using the chemical vapor deposition (CVD) technique with a particular organic molecular precursor 3,3-dimethyl-1-butene (C6H12). An exhaustive computational search of the potential energy surface reveals its unique sp2-sp3 hybrid bonding topology. Synergistic studies involving a large number of experimental techniques and multi-scale first-principles calculations reveal the origin of its novel properties due to the special arrangement of sp2 carbon atoms in lattice. The discovery of this U-carbon, named such because of its unusual structure and properties, can open a new chapter in carbon science. Carbon, the building block of life on Earth, is a unique element in the periodic table. Due to its flexible bonding characteristics categorized by sp3, sp2 and sp1 hybridization of the s and p- orbitals, it forms over ten million compounds whose properties are intimately related to their structures. Among these, diamond and graphite, two of the best-known three-dimensional (3D) 3 2 forms of carbon with sp and sp bonding, respectively, display strikingly different structure and properties. Since the discovery of C60 fullerene (1) we have witnessed the emergence of several new multi-dimensional carbon allotropes (2-5) exhibiting a range of spectacular properties. For decades, there has been a constant search for metallic and magnetic 3D carbon material with little success. The electrical conductivity (2~3×105 S/m) of graphite in its basal plane comes from the 2D graphene layer with zero density of states at the Fermi level. A similar case is found in the bundle of armchair nanotubes, where the conductivity is from the isolated 1D nanotube along the bundle axis (6). A cubic phase of carbon formed under 3 Tera-Pascal (3×1012 Pa) pressure was reported to be metallic, but loses its stability when the pressure is removed (7). None of these carbon materials is magnetic. Although carbon magnetism has been reported in amorphous carbon structures (5,8), other studies suggest that the defect-mediated magnetism in the non-crystalline carbon is paramagnetic or has weak magnetic ordering (9-10). While hydrocarbon (11) and functionalized graphene (12-13) can also become magnetic, no pure 1 crystalline magnetic carbon has ever been found. An early experiment claiming the observation of magnetic carbon was later found to be contaminated with magnetic metal impurities (14). It is expected that a 3D metallic and magnetic carbon would be metastable with energy higher than that of graphite and would belong to certain local minimum in the potential energy surface (PES). We hypothesized that metastable phases can be achieved by limiting the accessible region in the PES during synthesis and "forcing" the produced structure into prescribed states. One possible way to realize this is to use selected molecular precursors rather than individual atoms in the formation. Given that the existing bonding structure inside the chosen molecular precursor will encounter an energy cost for any bond breaking and rearrangement, it may be energetically preferable for the precursor to maintain certain original bonding features when forming the metastable structure, especially in rapid reactions. This, to some extent, would limit the accessible region in the PES and novel carbon phases could form (15-16). Fig. 1. Synthesis of U-carbon (UC). (A) Molecular structure of the precursor 3,3-dimethyl-1-butene (C6H12). (B) UC samples exhibit mirror-like appearance with metallic shine. We chose 3,3-dimethyl-1-butene (C6H12) (Fig. 1A) as the molecular precursor with the aim of forming a new metastable sp2-sp3 hybrid bonding system. Samples were synthesized in a hot-wall CVD reactor operating at atmospheric pressure and temperatures ranging from 700- 1000°C. The precursor 3,3-dimethyl-1-butene (C6H12) is added to the reactor via bubbling argon gas through it, upstream of the reactor. Injection of the precursor is started once the reactor has reached the desired temperature for growth (typically 800°C) and halted prior to cooling down. Growth was found to be substrate dependent, with metal oxides, copper, nickel, boron nitride and silicon oxide and nitride supporting growth. In contrast, un-oxidized silicon and glassy carbon do not show any indications of film growth (also see Section 1.1.1 in the Supplementary Information, SI). As a control, another form of hexane, cyclohexane with the identical chemical formula C6H12 but distinctively different molecular structure, was tested as precursor under the same growth conditions. From the measured FTIR and phonon Raman (Section 1.1.2 in SI), it is found that the resulting products from the two precursors are very different. This suggests that the precursor molecule's topology indeed plays a crucial role in the structure of the resulting carbon. 2 Thus, the synthesis supports a growth model where the hydrocarbon feedstock does not break down to atomic radicals and retains some of the original backbone structure. From the studies of thermal stability of hydrocarbons in the temperature range and flow rates comparable to the current settings of CVD growth, it is known that hexanes will not extensively break down (17-18). Another key distinction in the CVD here is the lack of hydrogen in the feedstock gas. Hydrocarbon cracking relies on hydrogen binding to the catalyst to lower the barrier energy for the hydrocarbon to undergo scission (19-21). The lack of large quantities of available hydrogen render such reaction pathway much less probable. With catalytic cracking suppressed, dehydrogenation becomes the dominant reaction at the catalyst surface (22), resulting in the formation of a carbon radical site on the molecule followed by bonding to the solid carbon. The synthesized carbon material (Fig. 1B), named U-carbon (UC) because of its unusual structures and properties (to be discussed later), shows a high reflectivity from far UV to mid IR (Section 1.1.3 in SI). Unlike the appearance of amorphous carbon samples, the U-carbon samples are macroscopically uniform in nature and shine like metals with ultra-smooth surfaces (Fig. S3 in SI). There is no hydrogen left in the sample (Section 1.1.4 in SI), suggesting complete dehydrogenation of the precursor molecules during the synthesis. The measured XRD (Fig. S5 and Section 1.1.5 in SI) exhibits clear and distinctive peaks throughout the region, indicating a crystalline nature of the sample. The unusual broadness yet high intensity of the first peak around 26° suggests that it should be contributed by multiple, rather than a single, sets of crystal planes. This could correspond to a group of structures due to different stacking configurations or level of binding, leading to slightly different d-spacing. No single crystalline carbon materials known can match the measured XRD (e.g. graphite or diamond as shown in Fig. S6 in SI). Thus, UC clearly represents a new form of carbon. We carried out a comprehensive structure search for UC and matched the simulated XRD with the experiment. One method applied is to optimize topologically assembled precursor units (Section 1.2.1 in SI), which resulted in a layered structure (Fig. 2A). Each layer, called U- graphene, has a buckled structure with three equally separated sublayers of carbon atoms (Fig. S8 in SI). Structures searched by an unbiased global optimization method based on individual 3 atoms (Section 1.2.1 in SI) resulted in a pure sp bonded structure (Fig. 2D). 3 Fig. 2. Possible configurations of U-carbon (UC). These include UC-aa, UC-aa', UC-aa'ab, UC-ab and UC-aa''. They originate from the possible stacking configurations of U-graphene (Fig. S2 in SI). In the UC-aa configuration, staggered-arranged sp2 carbon atoms are shown in blue and the sp3 carbon atoms in the unit cell are in red. The corresponding 3D views of the structures are shown with a tilted angle to illustrate the particular arrangement of the carbon atoms in the direction perpendicular to the paper. Note that the visually triangular arrangement of carbon (numbered as 1, 2 and 3) in the 2D view is actually composed by three carbon atoms in different planes parallel to the paper. This special geometrical arrangement is due to the uniquely buckled structure of U-graphene (Fig. S8 in SI), making the detected signals of the carbon atoms (1, 2 and 3 alike) overlapping with each other. Transitions between the structures can be realized by relative sliding and approaching of the neighboring layers, forming new bonds as demonstrated by the connections between the numbered atoms (4 and 5 connecting to 1 and 3). These configurations are shown along the calculated potential energy profiles of two identified reaction routes. The interconversion barriers separating UC-aa'ab, UC-ab and UC-aa'' from the others are higher than 0.4 eV/atom. We found that the layered and the sp3 bonded structures actually belong to the same carbon material, caused by different stacking sequence of U-graphene. The AA-stacking of U- graphene results in the layered configuration (UC-aa, Fig.